Do not index
Why did Steven Bartlett's team stop using AI to write his LinkedIn posts? Because readers could tell, and engagement dropped once they could. Business Insider reported that FlightStory, the company behind Bartlett's personal brand, pulled AI out of his LinkedIn writing entirely after audiences began spotting the automated posts. My read is blunt. The retreat from AI was correct. The tactic that replaced it is not. The team now leaves occasional spelling mistakes in on purpose to signal that a human wrote the post, and planted typos are not humanity. They are a performance of humanity, and audiences will spot the performance the same way they spotted the AI.
This one is for a specific reader. If you are a ghostwriter charging $5k to $30k per month to write in a founder's voice, a founder whose personal-brand content feeds real pipeline, or the content lead in a 3 person agency team deciding how much of the writing process to hand to a model, one of the biggest personal brands on the platform just reversed course in public, and the reasons behind that reversal are your market shifting under you.
This is not for everyone. If your plan is to keep a fully automated content pipeline and add typos as camouflage, this article will not change your model. Skip it too if you believe voice is a settings problem, something a better prompt will eventually solve. It is not, and the founders paying $30k a month for ghostwriting already know it is not.
The FlightStory quote worth sitting with came from Christiana Brenton, the company's director of business development. "AI mistakes are really noticeable. When the world wobbles to the left, opportunity lies to the right. What will stand out on LinkedIn now more than ever? Real words written by a human." That is according to Business Insider's July coverage. She is right about the opportunity. The open question is what real words written by a human actually look like on the page, because the answer is not misspellings.
What makes writing sound human on LinkedIn
I run a ghostwriting content team, so this is the question my work lives inside. After well over 200 posts a year across clients, the pattern is consistent enough that I gave it a name so my team can apply it before anything ships. I call it the Texture Test, and it is three questions asked of every draft. Could a stranger have written this? If yes, the draft has no specificity, no client numbers, no dates, no named process, nothing that had to be lived to be written. Would anyone push back on this? If no, the draft has no opinion, and text without a position is filler regardless of who typed it. Does it sound like the person talking? If not, the rhythm is off, because people write in uneven sentence lengths, start sentences with And, and circle back when something matters to them. AI text averages all of that away, which is exactly why it reads as detectable at scale.
A draft that passes all three questions reads human even when it is polished and typo-free. A draft that fails them reads synthetic even when a human wrote every word, which is the uncomfortable half of this story. Plenty of human-written LinkedIn content fails the Texture Test, and it was failing long before AI arrived. The models did not invent generic writing. They industrialized it.
Why planted typos backfire
Deliberate mistakes are a signal strategy, and signal strategies collapse the moment the audience learns the signal is manufactured. The first time a reader learns typos are being left in on purpose, every typo becomes evidence of the opposite of what it was planted to prove. Authenticity theater has a short shelf life because it trains your sharpest readers, the exact buyers you want, to look for the seams. Specificity does not have this failure mode. A real number from a real engagement cannot be reverse-engineered into a trick, which is why practitioner detail beats performance every time, and why I push founders toward practitioner-first positioning instead of thought leadership costume.
Bartlett's team reversing course tells you where the market is going. The audiences of 2026 have consumed enough machine text to develop palate fatigue, and they are recalibrating toward anything that reads lived rather than generated. For a founder between 5,000 and 50,000 followers, that recalibration is the opening. You do not need Bartlett's reach to benefit from it. You need an archive of posts that could only have come from your business, your numbers, and your positions. The teams that build that archive over the next year will own a kind of trust that no volume strategy can buy back later, because trust compounds and camouflage does not. The strategic question is not whether AI belongs anywhere in your process. It is whether anything in your published output could be mistaken for everyone else's, and what that resemblance is quietly costing you.
